Design and Management of Fluidized-Sand Bio Filters

Aquacultural Engineering 34 (2006) 275302 www.elsevier.
com/locate/aqua-online
Design and management of conventional uidized-sand biolters

Steven T. Summerfelt *
The Conservation Fund Freshwater Institute, 1098 Turner Road, Shepherdstown, WV 25443, USA
Abstract Fluidized-sand beds are an efcient, relatively compact, and cost-competitive technology for removing dissolved wastes from recirculating aquaculture systems, especially in relatively cool or coldwater applications that require maintaining consistently low levels of ammonia and nitrite. This paper describes several types of ow injection mechanisms used in commercial uidized-sand biolters and provides criteria for design of ow distribution mechanisms at the bottom of the uidized bed. This paper also summarizes the most critical aspects of sand selection, as well as methods for calculating or experimentally measuring uidization velocities and pressure drop for a given lter sand size distribution. Estimates of nitrication rate, ammonia removal efciency, carbon dioxide production, and oxygen consumption across uidized-sand biolters are also provided for various conditions. Fluidized-sand biolter operational and management practices are also described. # 2005 Elsevier B.V. All rights reserved.
Keywords: Fluidize; Biolter; Bioltration; Ammonia removal; Sand; Water recirculation or reuse; Aquaculture
1. Introduction Biolter selection inuences capital and operating costs of recirculating aquaculture systems, their water quality, and even the consistency of water treatment. A perfect biolter would remove all of the ammonia entering the unit, produce no nitrite, support dense microbial growth on an inexpensive support material that does not capture solids, require little or no water
* Tel.: +1 304 870 2211; fax: +1 304 870 2208. E-mail address: s.summerfelt@freshwaterinstitute.org.
pressure or maintenance, and require a small footprint. Unfortunately, no biolter type can meet all of these objectives, but each biolter type has there own advantages and limitations. In addition, different factors considered in biolter selection can shift in relative importance depending upon production system requirements. For example, in recirculating systems used to culture salmonids, which are species that are relatively sensitive to unionized ammonia- and nitrite-nitrogen, a biolters capacity to reliably maintain low levels of total ammonia-nitrogen and nitrite-nitrogen could be as important a consideration
0144-8609/$ see front matter # 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.aquaeng.2005.08.010
276
S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302
Nomenclature Ab Aorif A1 C Deq cross-sectional area of uidized bed column (cm2) area of orice (cm2 or m2) uidization constant,
3 e3 rrp rgcDeq 63 m2 1e2
D10
D50 D60 D90 g Hbed Horif L Le DP Qbiof Qorif Re1 Sb SGp SGw T UC vmf v0 Vb
orice discharge coefcient for sharpedged, submerged orices (0.6) equivalent diameter, diameter of a sphere with the same volume as the particle of media (cm) effective size, size of an opening which will pass only the smallest 10% of the granular media (cm) mean size, sieve size which will pass 50% of the granular media (cm) sieve size which will pass 60% of the granular media (cm) calculating size, sieve size which will pass 90% of the granular media (cm) gravity constant (980 cm/s2) headloss due to ow through a granular bed (cm of water) headloss due to ow through orice (cm of water) depth of loosely-packed (static) granular-media bed (cm) depth of expanded (uidized) granularmedia bed (cm) headloss across a bed of granular media, m of H2O ow rate of water through biolter (L/ min) ow rate of water through orice (cm3/s) uidization Reynolds number for rv0 cDeq expansion model, 6m1e bed specic surface area (cm1) specic gravity of the particle (unitless) specic gravity of water (1.0 unitless) temperature (8C) uniformity coefcient minimum uidization velocity (cm/s) uid supercial velocity (cm/s) volume of bed (cm3)
Greek letters e static bed porosity of a loose packed bed, i.e., void fraction (unitless) ee expanded bed void fraction (unitless) m uid viscosity (g/cm/s) r uid density (g/cm3) rp density of a particle of media (g/cm3) c sphericity, the ratio of the surface area of a sphere of equal volume to the actual surface area of the particle (unitless)
as the biolters capital and operating costs (Summerfelt et al., 2001). Conventional1 uidized-sand biolters (FSBs) have been widely adopted in North America, especially in recirculating systems that must reliably maintain excellent water quality to produce species such as salmon smolt (Forsythe and Hosler, 2002; Holder, 2002; Wilton, 2002), arctic char (Summerfelt and Wade, 1998; Summerfelt et al., 2004a), rainbow trout (Heinen et al., 1996; Summerfelt et al., 2004b), endangered sh (Montagne, 2004), and tropical or ornamental sh (Weaver, 2005). FSBs can typically remove 5090% of the ammonia each pass and thus maintain total ammonia-nitrogen and nitrite-nitrogen concentrations in their discharge of 0.10.5 mg/L and <0.10.3, respectively, in cold- and cool-water aquaculture systems (Heinen et al., 1996; Summerfelt et al., 2004b). FSBs can be less expensive and more compact than other biolter types (Table 1), even when they are sized to provide excess nitrication capacity (Summerfelt and Wade, 1998; Timmons et al., 2000). The cost of surface area in FSBs is low (i.e., $ 0.050.004 m2 surface area) because lter sand has a high specic surface area (i.e., 4000 20,000 m2/m3) and is low cost, approximately $ 70 200 m3 of sand delivered (Summerfelt et al., 2004b). Individual FSBs can treat both small or large ows, with single FSBs treating as much as 190 L/s of water ow. FSBs can be circular or rectangular in shape, can
1 Non-conventional uidized biolters use an expanded or moving bed media material other than sand, such as granular activated carbon, which is operated in an upow conguration, or various types of relatively small plastic media, which are operated in either an upow or a downow conguration that depends upon the specic gravity of the media.
S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302 Table 1 Comparison of xed costs associated with different biolters that were all sized to meet the ammonia removal capacity for a tilapia farm capable of producing 454 Mton (1,000,000 lb) annually (from Timmons et al., 2000) Biolter type Rotating biological contactor Trickling biolter Bead lter Conventional uidized-sand biolter Cyclo biolter Farm cost ($) 668000 620000 296000 124000 76000 Cost, US$ per kg/year 0.31 0.28 0.14 0.054 0.036
277
be contained within plastic, berglass, concrete, or enamel-coated steel tanks, and can be constructed by personnel on site. In addition, FSBs can be relatively easy to manage because they do not lter solids from the passing ow and the actively growing microbial biomass in the expanded bed can be readily harvested by siphoning the lightest, i.e., thickest and oldest, biolm coated particles from the top of the bed. On the down-side, FSBs are relatively complex to design. FSB do not aerate, as do trickling lters. Therefore, FSBs should always be designed with a cascade column placed immediately downstream to strip dissolved carbon dioxide and bring the dissolved oxygen up to near 90% of saturation. FSB must also be operated within a fairly narrow water ow range, i.e., within about 1030% of its design ow, in order to maintain proper bed expansion. If water ow through a FSB ceases for more than approximately 624 h, depending upon conditions, the static but biologically active bed can turn anaerobic, resulting in a signicant loss in nitrication capacity. Additionally, pumping water through a FSB requires water pressure to lift the sand bed (i.e., about 1.0 m of water head is required to expand every 1.0 m of static sand bed depth) and to overcome headloss designed into the FSBs ow inlet structure and any elevation difference between the water level in the pump sump and the top of the FSB (Summerfelt, 1996; Summerfelt et al., 2004b). Thus a total dynamic pumping head of 0.350.55 bar (5 8 psig), can be required to move water from the pump sump to the FSB overow, depending largely upon the height of the FSB (Heinen et al., 1996; Summerfelt et al., 2004a, 2004b). In typical recirculating system designs for salmonid production (Summerfelt et al., 2004a, 2004b), once the water has been pumped
through the FSB and exits the top of this vessel, the elevation achieved is used to gravity ow the water through the carbon dioxide stipping unit, low head oxygenator, culture tank, particle trap, and microscreen lter, returning to the pump sump. So the pumping energy through the FSB supplies approximately 90% of the mechanical power used within these recirculating systems, i.e., the remaining 10% of the mechanical power is used to ventilate the stripping columns and intermittently turn the drum lters. The purpose of this paper is to provide design and operation criteria for conventional uidized-sand biolters. Toward this end the paper provides (1) methods and criteria for obtaining uniform ow distribution in FSBs, (2) criteria for lter sand selection, (3) techniques for measuring bed expansion for a given lter sand, (4) calculation of headloss across expanded beds, (5) calculations of the minimum water velocity required to provide uidization and the bed expansion achieved at a given water velocity, (6) descriptions of the effects of biolm growth on uidization hydraulics and vertical stratication within the expanded bed, (7) estimates of nitrication rate and ammonia removal efciency, (8) estimates of dissolved carbon dioxide production and dissolved oxygen consumption, and (9) practices for operating and managing FSBs.
2. Mechanisms for ow injection Uniform water ow distribution at the base of the sand bed is critical for reliable operation of FSBs (Summerfelt and Cleasby, 1996; Summerfelt et al., 1996, 2004b). In addition to distributing an equal amount of ow across the base of the FSB, a properly designed ow distribution mechanism must also operate without detrimental fouling (or incorporate a mechanism to clear fouling from a plugged distribution system), prevent loss of lter sand, and support the sand bed in some designs. At least ve different ow distribution mechanisms have been used to uniformly inject water at the base of large FSBs in recirculating aquaculture systems. Four of the ve ow distribution mechanisms include a pipe manifold that originates at the top of the FSB and one or more vertical pipes that carry the ow down the inside of the
278
reactor to the base of the sand bed. Bringing ow into the FSB from above avoids having to penetrate the vessels wall(s) with distribution pipes and also prevents the hydraulic head of water in the vessel from back owing through the distribution piping when a siphon break or check valve has been installed. The only ow distribution mechanism for FSBs that is commonly used in recirculating aquaculture systems that does not utilize a manifold at its top is the CycloBio1 FSB, which is discussed in detail below. The CycloBio1 FSB is also only supplied in a cylindrical vessel, whereas all of the other types of FSBs can be constructed with a cross-sectional area formed out of most any shape. 2.1. Gravel covered horizontal pipe or false oor manifold One ow distribution system for FSBs consists of either a gravel covered pipe-manifold or false-oor distribution chamber that is sometimes used in relatively small recirculated systems (Malone and Burden, 1988) or in large-scale wastewater treatment systems (Jewell, 1990; Cooper and Atkinson, 1981; Sutton and Mishra, 1991). In some instances, one to four layers (each about 7.6 cm deep) of graded gravel are leveled over the distribution plate or piping, with the coarsest gravel next to the distribution plate or piping (Cleasby, 1990). However, high water velocities in the vicinity of the inlet orices can cause gravel movement, particularly if the water velocity is not gradually increased during initial bed expansion (Cleasby, 1990). In addition, the gravel layers are not uidized under normal operating conditions and, thus, are susceptible to plugging from solids entrapment and from microbiological growth. In order to avoid the use of layered gravel and its associated problems, industrial and municipal wastewater treatment applications have used nozzle-type ow distributors in combination with the false oor type ow distribution mechanism (Sutton and Mishra, 1991). However, when the distribution nozzle openings are small enough to exclude the majority of sand, there is an increased likelihood that the nozzles will plug or foul (Sutton and Mishra, 1991). In addition, their relatively high cost and proprietary nature has probably limited the application of more nozzle-type ow distribution systems that originated in the wastewater treatment industry.
2.2. Vertical pipe manifold Possibly the rst type of ow distribution mechanism developed specically for application in recirculating aquaculture systems incorporates a pipemanifold, originating at the top of the vessel, and then distributes the ow into vertical pipes that extend down to the base of the sand bed (Fig. 1) (Weaver, 1991, 2005). The vertical injection pipes are equally spaced across the plan area of the FSB and transport water ow to near the oor of the vessel where orices in each probe uniformly distributes the ow directly into the sand (Fig. 1), without using layered gravel. If on occasion a ow injection probe plugs with sand, a mechanism is provided to ush the blockage from the probe. A uidized-sand bed biolter of this type, designed by Dr. Dallas Weaver (Scientic Hatcheries, Huntington Beach, CA), was used successfully at The Conservation Fund Freshwater Institute (Shepherdstown, WV) in the early 1990s (Heinen et al., 1996). This FSB design is marketed through Aquaneering Inc. (San Diego, CA) and has been widely applied in many types of recirculating aquaculture situations, but especially in applications that require ne sand FSBs to maintain high quality water (Weaver, 2005). 2.3. False oor orice distribution plate A second distribution mechanism specically developed for application in recirculating aquaculture systems consists of a pipe manifold system that originates above the vessel and connects to a distribution chamber below a false-oor supporting the sand bed at the base of the vessel (Fig. 1). The central vertical pipe manifold is branched into four pipes, positioned in an H pattern when viewed from above, just before the four pipes connect to the falseoor. The false oor contains equally spaced orices to uniformly distribute the water ow across the crosssectional area at the base of the sand bed. The geometry of the distribution chamber, the location of the vertical laterals, and the spacing and size of the orices are dependent upon the ow of water and on the diameter of sand (Eric Swanson, Maritime Aqua Service Ltd., Northeast Harbour, Maine, personal communication, 1993). A FSB of this type can be purchased from Legay Fiberglass Limited (Waverley, NS, Canada).
279
2.4. Horizontal pipe manifold Another ow distribution mechanism, the horizontal pipe manifold system (Fig. 1), was also developed specically for application in FSBs for
recirculating aquaculture systems (Summerfelt, 1996; Summerfelt et al., 1996). Design of the horizontal pipe manifold ow distribution mechanism was based on the pipe lateral system used to backwash gravity sand lters that are typically found in municipal drinking
Fig. 1. Four of the most common ow distribution manifold systems used in conventional uidized-sand biolters for recirculating aquaculture systems include: (A) the vertical pipe manifold, (B) the horizontal pipe manifold, (C) the false oor orice distribution plate, and (D) the CycloBio1 w/slotted inlet manifold.
280
Fig. 1. (Continued ).
water and wastewater applications. However, the ow distribution system was modied so that the manifold originated and terminated at the top of the biolter. The overhead manifold pipes distribute the ow to
equally spaced vertical pipes that run down the inside wall of the vessel to its base (Fig. 1). At the base of the vessel, each vertical pipe elbows 908 and runs horizontally across the oor to the opposite wall
281
where it again elbows 908 to run up the wall to the top of the vessel (Fig. 1). Flow distribution orices are located on the section of horizontal pipe running along the oor of the vessel (Fig. 1). This design places threaded caps or valves at the top and end of each distribution pipe (Fig. 1), which can be temporarily opened as necessary to ush sand or debris from individual distribution pipes that have plugged (Summerfelt et al., 1996). If even more vigorous action was required to unplug a lateral, a garden hose or wire-rooter can be run down through the top of the tee ttings (the threaded cap covered opening) to ush debris. This design also allows for installation of control valves between the manifold pipe and individual distribution pipes (Fig. 1), which allows more ow to be forced through individual pipes to ush out sand or other debris when necessary. A swing-ex type check valve, which is located between the pumps and pipe manifold located above the biolter vessel, prevents the hydraulic head of water in the vessel from back owing through the distribution piping when pump pressure is lost and the expanded bed is deuidizing. Alternatively, a reissued US Patent by Goldman and Rosenau (2000) claims a method where at least one channel or reservoir of uid is provided over at least a selected portion of the distribution network, with uid from such column/reserve being released with a vacuum breaker in order to maintain ow and pressure in the distribution network as the media bed is deuidized. The horizontal pipe lateral FSB design has been applied in numerous coldwater recirculating aquaculture system applications in North America (Summerfelt et al., 1996, 2004a; Summerfelt and Wade, 1998; Forsythe and Hosler, 2002; Holder, 2002) and has been marketed by PRAqua Technologies (Nanaimo, BC, Canada). Criteria for calculating the size and separation of pipe laterals and ow injection orices for the horizontal pipe lateral type ow injection mechanisms are well understood (Weber, 1972; Montgomery, 1985; AWWA, 1990; Summerfelt, 1996). Flow injection orices and pipe laterals are typically separated at xed intervals of between 7.5 and 30 cm (312 in.). The size and number of ow injection orices are selected to provide a constant and controlling loss of head at each orice to produce
an equal ow through all orices. Montgomery (1985) recommends sizing the distribution orices to create an orice headloss of at least 0.6 m. However, an orice headloss of just greater than the headloss across the expanded bed can be expected to provide equal ow distribution, assuming that the distribution system is sized so that the ow velocity within the pipes are reasonably low and uniform throughout the lter area (Weber, 1972; Montgomery, 1985; AWWA, 1990). The headloss across an orice of a given diameter at a given water ow rate can be estimated from the following equation: Qorif 2 1 Horif (1) 2g CAorif where Horif is the headloss due to ow through orice (m of water); Qorif the ow rate of water through orice (m3/s); Aorif the area of orice (m2); g the gravity constant (9.81 m/s2); C the orice discharge coefcient for sharp-edged, submerged orices (0.6). In practice, ow distribution orices of approximately 6.412.7 mm (0.250.5 in.) diameter are recommended (Weber, 1972; Montgomery, 1985; AWWA, 1990). The ow distribution orices should be aligned in two rows located on opposite sides of each horizontal pipe, running the length of the horizontal pipe, and directed downward so as to dissipate the energy of the water jets (Weber, 1972). Additional ratios are provided by Weber (1972) to act as guidelines to size the orices, pipe laterals, and pipe manifold in order to obtain uniform ow distribution, i.e., total area of orifices : cross-sectional area of bed 0:0015 to 0:005 : 1 cross-sectional area of pipe-lateral : total area of orifices served 2 to 4 : 1 cross-sectional area of manifold : total area of pipe-laterals served 1:5 to 3 : 1 (4) Use of these ratios during the design of the ow distribution mechanism will help to (a) select pipe manifold and lateral sizes that produce water velocities that are reasonably low and uniform throughout the entire lter area, and (b) provide injection orice (3)
(2)
282
velocities suitable for introducing a controlling headloss (Montgomery, 1985). Creating a headloss across the orice that is much lower than the head required to expand the sand bed would not provide uniform ow distribution under the sand bed, but rather it would produce water spouting through regions of the bed while other regions of the bed remain static. In contrast, an orice headloss that is signicantly larger than the head required to expand the sand bed would increase pumping costs and also produce a jetting action that would be more likely to damage the base of the vessel and abrade the sand in the jetting zone. Water jets emitting from the downward facing orices (at an angle 458 below horizontal) have been known to sand blast holes through a 6 mm thick berglass oor of a FSB within only 7 days of start-up (Summerfelt et al., 1996). The AWWA (1971) has also reported on jet action producing problems at the base of lter beds. Therefore, to reduce the likelihood of the nozzle created jet action from sand-blasting through the vessel wall or oor, an abrasion resistant concrete pad or brick liner should be constructed on the biolter oor, approximately 1015 cm below the distribution pipes, and around the bottom 1020 cm of the sides of the FRP vessel (Summerfelt et al., 1996). 2.5. CycloBio1 with slotted inlet manifold The CycloBio1 FSB was developed by Neil Helwig of Marine Biotech Inc. (Beverly, MA) specically for application in recirculating aquaculture systems, but it uses a ow distribution mechanism that is radically different from that used by the vertical pipe manifold, horizontal pipe-manifold, and falseoor ow distribution mechanisms (Fig. 1) (Timmons et al., 2000; Summerfelt et al., 2001, 2004b). The CycloBio1 FSB injects water tangentially into an annular space that surrounds the base of the circular vessel and is integrated with the vessel wall. This continuous tangential injection of water through the annular space creates strong water rotation within the annular chamber and also forces water to enter the FSB vessel through a slot at the base of the sand bed and around the circumference of the vessel (Fig. 1). An inverted cone, incorporated into the center of the oor of the vessel (Fig. 1), is used to increase the upwardowing water velocity at the base of the sand bed,
which helps to improve sand bed expansion. Water ow injection within a CycloBio1 FSB has some analogies with the slotted inlet design that is used to uniformly introduce air within agriculture livestock buildings (Timmons et al., 2000). However, the strong rotating ow created within the annular space of a CycloBio1 FSB also imparts a cyclonic rotation within the expanded sand bed when clean sand bed expansion exceeds approximately 6080%. The water velocity through the CycloBio1 inlet slot is relatively small compared to the water velocity passing through the ow injection orices on horizontal pipe lateral ow distribution mechanisms. Therefore, the CycloBio1 FSB operates with relatively little headloss across the ow inlet slot (i.e., approximately 0.2 0.4 psig), which is approximately 1.52.0 psig less than the orice headloss designed into horizontal pipe manifold ow distribution mechanisms (Summerfelt et al., 2004b). Cyclonic rotation of the sand bed is a signicant force contributing to uniform bed expansion in a CycloBio1 FSB (Summerfelt et al., 2004b). In addition, CycloBio1 FSBs appear to operate with more uniform bed expansion when their expanded bed depth is ! twice the diameter of the vessel (Summerfelt et al., 2004b). Under these conditions, CycloBio1 FSBs have been found to be simple to operate and to re-uidize after shut down (Summerfelt et al., 2004b). However, if a check valve failure or other unexpected event allows water to backow through the CycloBio1 vessels tangential inlet, the plugged annular chamber that results can be rapidly cleared of packed sand by removing the blind ange covering an access port into the annular chamber and briey turning on the pump supplying the FSB to ush sand from the inlet area. Experience at TCFFI has shown that when the blind ange has been reinstalled over the access cover, the CycloBio1 FSB can re-uidize and clear the remaining sand from the annular chamber. Note that the blind ange over the access port into the annular chamber must not be removed until after water inside the vessel has been removed. In addition, no oor abrasion below the slotted inlet has been observed in the two large-scale CycloBio1 FSBs evaluated at TCFFI, probably because the water velocities at the inlet slot are relatively low and the water is injected parallel to the vessel oor (Summerfelt et al., 2004b).
283
Fig. 2. Sieve analysis of three lter sands that were evaluated in uidized-sand biolters at the Conservation Fund Freshwater Institute. The sands D10, D60, and D90 are the sieve opening sizes that would pass only the smallest 10, 60, or 90% of the granular sample by weight. D10 is the sands effective size.
CycloBio1 FSBs can be purchased from Marine Biotech Inc. All of the ow distribution mechanisms described above, except those mechanisms that rely on layered gravel or inlet nozzle structures, have typically worked well in FSBs used in recirculating aquaculture systems. However, as in the wastewater treatment industry (Sutton and Mishra, 1991), the gravelcovered and nozzle structure covered distribution systems have had some problems with plugging and channeling in large lters.
3. Sand selection criteria FSBs use extremely hard, whole grain, nely graded crystalline silica sand, which has a specic gravity of 2.65. Silica lter sand suppliers can be located by searching the World Wide Web or a list of lter sand suppliers such as is provided by AWWA (2004). Filter sands are typically pre-sieved to produce a distinct size range, which is usually specied by an effective size and uniformity coefcient (Cleasby, 1990). ASTM (1985) standard test procedure C136-
84a should be used for conducting sieve analysis of a granular sample. The sands effective size (D10) is the sieve opening size that would pass only the smallest 10% of the granular sample by weight, as read from the log-probability plot of the sieve analysis for a given sand sample (Fig. 2). The sands uniformity coefcient (UC) is a quantitative measure of the variation in particle size within the sand sample that is dened as the ratio of D60:D10. The sands D60 is the size for which 60% of the sand is smaller by weight, as read from the same log-probability plot of the sieve analysis results (Fig. 2). The D10 is the fraction of sand that will expand2 the most at a given supercial velocity in a FSB. The D90, an estimate of the sand diameter of the largest 10% of sand in the sample, is the sand fraction that will expand the least at a given supercial velocity. The D90 can also be read from the log-probability plot of
When a sand bed is expanded, its total height is some percentage greater than its initial static height. Percent bed expansion is calculated by subtracting the expanded bed height from the static bed height, dividing this difference by the static bed height, and then multiplying by 100.
2
284
the sieve analysis for the sand in question (Fig. 2). Alternatively, the D90 can be estimated if the D10 and UC of the sand are supplied with the use of the following equation (Cleasby, 1990): D90 D10 101:67 logUC (5)
The size and UC of lter sand are both critical for successful application in FSBs. Sand grains with relatively larger diameters will generally migrate towards the bottom of a FSB where they expand less than the relatively smaller sands that have migrated to the top of the bed. For a given graded sand placed in a FSB, the D90 fraction of the sand must expand at least 1020% at the design supercial velocity in order to minimize occurrence of static sand piles at the base of the bed. At the same time, the D10 fraction of sand must not expand excessively (e.g., over 150200%) to prevent it from washing out the top of the FSB (Summerfelt and Cleasby, 1996). Sands used in FSBs for recirculating aquaculture systems may have a D10 as small as 0.1 mm or as large as 1.0 mm, as well as a uniformity coefcient ranging from 1.3 to 1.8. A smaller UC is more ideal for lter sands, as this represents a smaller variation in the sand size distribution. During FSB design, a lter sand is selected that will provide a given D10 and size distribution that produces a mean clean-sand bed expansions of typically between 40 and 100% at a given supercial water velocity (Summerfelt and Cleasby, 1996). Final expansion of a FSB established with biolm may reach 200300%. Unfortunately, proper lter sand selection can be somewhat challenging. For this reason, during FSB design the sand bed expansion for a given sand source should be estimated at various supercial water velocities using both experimental test column studies and empirical calculations based upon the sands D10, D50, and D90. 3.1. Calculating pressure drop Water injected into the base of a FSB will ow up through void spaces between sand granules in the initially static bed. Water owing through the sand bed must overcome both viscous and inertial forces (Ergun, 1952), which combine to create an overall pressure drop across the static bed that increases with increasing water ow (Fig. 3). The sand bed begins to
Fig. 3. Illustration of the hydraulics of ow through a bed of granular media at supercial velocities (Q/Ab) above and below the minimum uidization velocity vmf (as shown by Fan, 1981). The vmf is the supercial velocity at the point of incipient uidization.
expand when the water velocity through the sand void spaces is sufcient to create a pressure loss that is greater than the apparent weight (actual weight less buoyancy) per unit cross-sectional area of the bed (Denn, 1980; Cleasby, 1990). When the bed is uidized, the individual sand granules are freely supported and tumble with the ow of water. The relative amount of sand bed expansion is dependent upon the density, shape and diameter of the particles, and the velocity of water (as discussed in the next section). Once the bed has been uidized, the pressure drop across the bed remains constant at all bed expansions at up to about 90% porosity (Cleasby, 1990), as diagramed in Fig. 3. When a sand bed is expanded, the net gravitational and buoyant force acting on the sand bed, i.e., the mass of the sand granules (rpgVb{1 e}) minus the mass of the liquid that they displace (rgVb{1 e}), is equal to the net upward force on the bed (DPAb), where (1 e) is the fraction of bed volume occupied by sand granules 0.530.58 (unitless), e the void fraction (i.e., porosity) of the loose static bed 0.420.47 (unitless), r the density of water (1.0 g/cm3), rp the density of sand (2.65 g/cm3), g the force of gravity (980 cm/s2); Ab the cross-sectional area of the bed in
285
cm2; Vb the volume of the bed in cm3, and DP the pressure drop across the uidized bed in dyne/cm2. Recognizing that the DP across the uidized bed is equal to the product of the headloss across the bed (Hbed, in cm of water), g, and r, then simplifying the force balance will produce the constant headloss equation for expanded beds (Denn, 1980): Hbed rp r 1 e SGp SGw 1 e r L (6)
the D10, D50, D90, UC, rp, e, and c for a given sand are obtained or assumed, these values can be used to calculate the supercial water velocity required to achieve a given bed expansion at a given water temperature, as will be discussed below. 3.2.1. Minimum uidization velocity Wen and Yu (1966) developed an equation to predict the minimum uidization velocity vmf at the point of incipient uidization of the sand, which only requires knowledge of the equivalent diameter of the sand granule (Deq), rp, r and uid viscosity (m): rrp rg 0:5 m vmf 33:72 0:0408D3 eq rDeq m2 33:7m (7) rDeq 3.2.2. Bed expansion versus velocity The porosity of an expanded bed (ee), which can never exceed 1.0 as the expanded bed depth goes to innity, can be calculated from e, L, and the expanded bed depth (Le), e.g. (Weber, 1972): ee 1 1 e L Le (8)
where L is the depth of the static bed in cm, SGp the specic gravity of the particle (unitless), and SGw is the specic gravity of water (unitless). This equation can be used to calculate the headloss per unit static bed depth, i.e., h/L, for an expanded bed of granular particles. Based upon this equation, a uidized bed of sand would require approximately 0.870.98 m of water head for every 1.0 m of initially static sand depth, if a static bed porosity of 0.420.47 is assumed (Summerfelt and Cleasby, 1996). Note from Eq. (6) that the headloss across an expanded bed is independent of sand size. 3.2. Calculating bed expansion as a function of water velocity and sand size Expansion of a clean sand bed can be estimated for a given supercial water velocity if the sands diameter, UC, rp, e, and sphericity (c) are known as well as the uids r and viscosity (m), both of which are temperature dependent (Wen and Yu, 1966; Dharmarajah and Cleasby, 1986). As described in Table 2, these sand characteristics can be determined from preliminary laboratory uidization studies and sieve analyses on samples of the same sand. These studies require facilities and time, not only to perform the studies but also to obtain the sand samples. Alternatively, the rp, e, and c of many sand sources can typically be assumed to be within the range of 2.6452.655, 0.420.47, and 0.70.8, respectively (Cleasby, 1990), for a loosely packed sand. The D10, D50, D90, and UC of each lter sand supply must still be determined from a sieve analysis (Fig. 2), but the sieve analysis can be obtained from the supplier. After logA1
The ee can also be predicted for a given supercial velocity v0 and a given sand Deq, rp, c and e, as well as the waters r and m using a phenomenological model developed by Dharmarajah and Cleasby (1986). The model of Dharmarajah and Cleasby (1986) results in a plot of two dimensionless numbers, A1 and Re1: A1 rrp rgcDeq 3 ; 63 m 2 1 e2 rv0 cDeq Re1 6m1 e e3
(9)
Dharmarajah and Cleasby (1986) t the curve resulting from the plot of A1 versus Re1 using a step-wise regression procedure, and report the polynomial relationship: (10)
0:56543 1:09348 logRe1 0:17971logRe1 2 0:00392logRe1 4 1:5logc2
286
Table 2 Calculations of media characteristics (from Summerfelt and Cleasby, 1996) Property Static bed porosity, e Denition and method of determination Denition. The porosity of a static bed of granular media is the fraction of the volume within the bed which is not occupied by particles: e volume of voids. The porosity required for uidization volume of bed calculations is the porosity of the static bed after it has been fully expanded (200%) and has just been brought to rest by gradually cutting back the hydraulic loading rate until the ow is zero Determination. The static bed void fraction of the sample of clean sand can be calculated from the total mass of the clean sand sample, the total volume of the sand sample, and the density of sand: e 1 total mass of sand=particle density of sand volume of bed Denition. The density of a particle is the ratio of particle mass to particle volume including its pore volume but excluding inter-particle voids Determination. The density of a particle can be determined for porous and non-porous media Non-porous media: The density of non-porous media can be estimated by quantifying the volume displaced by a given mass of particles Porous media: Geldart (1990) described methods for calculating the density of porous particles Denition. The density of the bed is ratio of bed mass (dry) to bed volume Determination. The density of the bed of granular media can be found by measuring the mass of the media and dividing by the volume that mass occupies Denition. The particle size distribution is dened as the relative percentage by weight of grains of each the different size fractions represented in the sample Determination. The size distribution of a sample of sand is generally determined from a sieve analysis and results in a table of the percentage of media ner than a given opening size. A plot on log-probability graph paper of the percent ner (normal scale) vs. the corresponding sieve size (log scale) will show a straight line for most natural sands (Weber, 1972) Denition. The effective size is dened as the opening size that will pass only the smallest 10%, by weight, of the granular sample Determination. The D10 can be taken from a log-probability plot of the particle size distribution Denition. The calculating size is the sieve size for which 90% of the grains by weight are smaller. The D90 provides an estimate of the largest sand in the sample and is the value used during design to calculate the velocity required to uidize even the largest sand Determination. The D90 can be determined from the particle size distribution as plotted on log-probability paper or can be approximated (Cleasby, 1990) by: D90 D10 101:67 logUC Denition. The uniformity coefcient is a quantitative measure of the variation in particle size of a given media and is dened as the ratio of D60:D10 Determination. The UC for a given granular media equals the D60 divided by the D10 values which are determined after plotting the results of a sieve analysis on log-probability paper Denition. The equivalent diameter of an irregular particle is dened as the diameter of a sphere with the same volume as the particle Determination. The average equivalent diameter of a sample of sand can be found by determining the mass of the average grain. The equivalent diameter of the sand can be calculated 1=3 grain 6 from: Deq p average mass of onesand particle density of Denition. The particle specic surface area is dened as the surface area per unit of particle volume. The Sp for particles which are perfectly spherical can be calculated
4pR 3 6 from: Sp sphere sphere surface area 4=3pR3 R D sphere volume
2
Particle density, rp
Bulk density of the bed, rb
Particle size distribution
Effective size, D10
Calculating size, D90
Uniformity coefcient, UC
Equivalent diameter, Deq
Particle specic surface area, Sp
(where R and D are the radius and diameter of the sphere, respectively). The Sp for particles which are not completely uniform must take into account the particles shape (e.g. sphericity, c, 6 dened below): Sp particle surface area cDeq particle volume
S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302 Table 2 (Continued ) Property Denition and method of determination
287
Determination. If c is known, or can be assumed, then Sp can be calculated from Deq and c using the equation directly above. If c is not simply assumed, the Sp can be determined from experiments measuring the headloss vs. ow rate across a loosely packed bed of granular media. A relationship between headloss and ow rate, developed by Ergun (1952), can be used for plotting granular media headloss data and solving for Sp according to the slope of the resulting line Bed specic surface area, Sb Denition. The bed specic surface area is the total surface area of particles per unit of bed volume Determination. The Sb is dependent upon how tightly the bed is packed or conversely on how much the bed is expanded. The Sb can be calculated for a given e once Sp is known, e.g.: Sb = Sp(1 e)
These equations can be solved to accurately predict ee for a given v0 and specic sand characteristics (Dharmarajah and Cleasby, 1986). However, the solution required to estimate ee is iterative. Summerfelt and Cleasby (1996) have applied the Dharmarajah and Cleasby (1986) model to graphically relate the v0 required to achieve bed expansions of 0, 50, 100, and 150% for sand Deq sizes that range from 0.05 to 1.5 mm in freshwater, i.e., 0 ppt salinity, at 25 8C (Fig. 4). Note that the v0 required to achieve a given ee for a given Deq also depends upon the waters r and m, which are both dependent upon the temperature and
salinity of the water. For a given sand at a given v0 , sand bed expansion decreases with increasing temperature (Fig. 5), largely due to the corresponding decrease in m. A large change in water temperature appears to create a larger affect on bed expansion than will a change from freshwater to nearly full-strength seawater (Fig. 5). Changing from freshwater (i.e., 0 ppt salinity) to seawater (i.e., 32 ppt salinity) will only cause a slight increase in overall sand bed expansion for a given Deq at a given water temperature (Fig. 5). Fig. 4 can be used to roughly estimate the expansion of the D10, D50, and D90 size classes for
Fig. 4. Relationship between uid supercial velocity v0 and bed expansion for sand of a uniform size (Deq) assuming typical values for c (0.75), e (0.45) and T (25 8C) (from Summerfelt and Cleasby, 1996).
288
Fig. 5. Relationship between bed expansion and supercial velocity measured at water temperatures of 4.4 8C (~) and 26.7 8C (&) and at water salinities of 0 ppt () and 32 ppt (- - -) for a 20/40 sand expanded in a 0.61 m diameter CycloBio1 containing 0.91 m depth of initially loose static sand. Data courtesy of Thomas Lauttenbach, Marine Biotech Inc., Beverly, MA.
a given lter sand at a given v0 . The bed expansion at a given v0 should be calculated for each D10, D50, and D90 size classes of a specic lter sand, because the larger diameter sand fractions move to the bottom of a FSB where they are expanded to a much lower extent than the smaller sand fractions that have migrated to the top of the bed. For example, according to the sieve analysis reported in Fig. 2, the US Silica #1 Dry Mapleton sand has a D10, D50, and D90 of approximately 0.19, 0.28, and 0.40 mm, respectively, and these sand fractions would expand approximately 150, 80, and 35%, respectively, at a v0 of 0.75 cm/s according to the bed expansion estimates provided in Fig. 4. In this example, it appears that the largest 10% of the sand would be expected to expand, whereas the nest 1020% of the sand could be siphoned out of the FSB immediately after its initial bed expansion to reduce its likelihood of washing out when the biolm becomes established. If the nest 1020% of the sand was not removed, then the overall bed expansion of the US Silica #1 Dry Mapleton sand would be expected to be approximately 80% (i.e., the bed expansion at D50 = 0.28 mm), which is in approximate agreement with data collected from 10 cm diameter test column studies conducted at TCFFI (Summerfelt, unpublished data). 3.3. Experimental technique to measure bed expansion from a sample of lter sand The author recommends that after the v0 and ee have been estimated for a given lter sand, but before
the recirculating aquaculture system design has been completed, that a sample of the specic lter sand under consideration be obtained and tested hydraulically to develop the most accurate estimate of the actual relationship between v0 and ee. In order to conduct bed expansion tests, order at least 814 L (0.30.5 ft3) of a representative sample of the lter sand from the desired supplier. This sample will mass approximately 1221 kg, because lter sand has a bulk density of approximately 1600 kg/m3 (100 lb/ft3). Fabricate a test column set-up that consists of a 10 cm nominal inside diameter clear PVC pipe with an overall height of approximately 2.7 m (9 ft). The test column should contain a 2.5 cm (1 in.) diameter inlet pipe located at the bottom end of the test column and a 2.5 cm (1 in.) diameter outlet pipe that is located within 30 cm (1 ft) of the columns open top. At the base of the column, the inlet pipe should be covered with at least 30 cm (1 ft) of layered gravel to distribute the ow under the sand for the short duration of the expansion test. A pressurized water line with a throttling valve that can supply anywhere from 0 to 16 L/min of water should be connected to the inlet pipe of the test column. A check valve should be located on the inlet water supply pipe, just before the supply pipe enters the test column, in order to prevent water from back owing out the bottom of the test column. The outlet pipe near the top of the test column should be plumbed to discharge to a drain, but should also allow for bucket testing of the discharge ow rate. Approximately 1 m of sand
289
should be added to the test column. Sand should be poured into the top of the test column after the test column contains at least 1 m of water. Before the sand bed expansion tests begin, slowly open the throttling valve on the water supply until the sand bed in the test column has expanded approximately 0.6 m. Allow the test column to ush out ne particles for approximately 1030 min. At this time, the top 10 20% of the expanded bed could be siphoned out if this will be the normal start-up procedure for the fullscale FSB. The depth of sand siphoned from the top of the expanded bed will depend largely upon the uniformity coefcient of the sand; siphoning is not required when the sands UC approaches 1.3. The water owing through the test column should be turned off slowly and the sand bed should be allowed to settle for a period of 30 min before the total depth of the loose packed static sand bed is measured. During this period of bed collapse, do not tap or vibrate the column to obtain more compression of the static sand. After recording the loose packed static bed depth, slowly turn up the water ow rate through the bed until the bed expands approximately 20% and then record the overall bed height at the corresponding water ow rate that has been determined using a bucket test or a calibrated ow meter. Repeat this test while increasing the water ow rates to increase the bed expansion in increments of approximately 20% (e.g., 20, 40, 60, 80, 100, and 120% bed expansions) until overall bed expansion equals or exceeds 120%. At this point all of the experimental data will be collected and the test column can be cleaned out and prepared for the next sand trial. The supercial water velocity (v0 ) encountered during each of the test conditions can now be calculated from each of the water ow rates recorded (Qbiof, L/min), i.e., by dividing each ow rate by the cross-sectional area (Ab) of the column: Qbiof L=min 1000 cm3 1 min v0 Ab cm2 1 L 60 s (11)
Table 3 Water velocities required to expand four different sands approximately 20, 50, 100, and 150% Sands tested Retaining sieve mesh sizes Effective size, D10 (mm) Uniformity coefcient D50 (mm) Velocity requirements 20% expansion 50% expansion 100% expansion 150% expansion 40/70 0.24 1.8 0.37 (cm/s) 0.5/0.4 1.0/0.8 1.4/1.4 1.9/1.9 30/50 0.45 1.4 0.59 0.7/0.9 1.3/1.5 2.0/2.4 2.7/3.1 20/40 0.60 1.4 0.79 0.8/1.4 1.9/2.2 3.1/3.3 4.1/4.2 18/30 0.80 1.3 0.99 1.3/1.9 2.7/2.9 4.6/4.2 5.9/5.2
The four sands tested had a D10 of 0.24, 0.45, 0.60, and 0.80 mm. The rst v0 reported is an average of measurements made during uidization tests in a 10 cm diameter test column. The second v0 was predicted for the mean sand size (D50) using the Dharmarajah and Cleasby (1986) model, assuming that e and c are 0.45 and 0.75, respectively, and a water temperature of 25 8C.
and (10), assuming e and c of 0.45 and 0.75, respectively, a water temperature of 25 8C, and D50 for each sand of 0.37, 0.59, 0.79, and 0.99 mm. Comparing the two v0 reported in Table 3, for a given sand at a given sand bed expansion, indicates that the Dharmarajah and Cleasby (1986) model can be used to estimate sand bed expansion. However, the assumption that each of the sand samples was represented by a e and c of 0.45 and 0.75, respectively, and a D50 did create some error in estimating the v0 requirements for a given sand (Table 3). Therefore, conducting the sand bed expansion tests in a test column is still deemed necessary. 3.4. Comparing sand bed expansion measured in test columns with full-scale FSB tests Over the last decade, three of the ve ow distribution mechanisms illustrated in Fig. 1 were evaluated at TCFFI in Shepherdstown, West Virginia, i.e., the vertical probe system (Heinen et al., 1996; Weaver, 2005), the horizontal pipe manifold (Summerfelt et al., 1996), and the CycloBio1 FSB (Summerfelt et al., 2004b). The full-scale vertical probe FSB and the horizontal pipe manifold FSB both appeared to produce a sand bed expansion that were consistent with or slightly lower (within approximately
The v0 measured in test columns at sand bed expansions of 20, 50, 100, and 150% are reported in Table 3 for four lter sand samples with a D10 of approximately 0.24, 0.45, 0.60, and 0.80 mm. Table 3 also contains estimates of v0 that were calculated using the Dharmarajah and Cleasby (1986) model, i.e., Eqs. (9)
290
10%) than would be predicted by the expansion test column results. The slightly lower bed expansions that resulted could be explained by the occurrence of small sand mounds about the perimeter of the bed, which stored a small volume of sand that was not uidized. However, when sand bed expansion within full-scale (2.7 m diameter) CycloBio1 FSBs was compared with sand expansion tests that had been conducted within a 10 cm test column (Fig. 6), results indicate that a given sand expands roughly 1040% less in a full-scale CycloBio1 vessel than the same sand was found to expand in a 10 cm diameter test column for a given hydraulic loading rate (Summerfelt et al., 2004b). The nest sand tested, the #1 Mapleton sand (D10 = 0.18 mm) from US Silica, had the closest match for bed expansion when comparing the test column to the full-scale CycloBio1 data. The consistently lower bed expansion in the full-scale CycloBio1 FSB was probably due to water spouting along the wall of the vessel and to increased formation of transient sand mounds that were most prevalent when overall bed expansion was less than approximately 50% (Summerfelt et al., 2004b). Overall sand bed expansion within the full-scale CycloBio1 FSB was closest to the sand bed expansion within the 10 cm diameter test column when overall bed expansion exceeded approximately 60% (Fig. 6). Because of these ndings, during the design of a full-scale CycloBio1 FSB, a given lter sand should be expected to expand approximately 10 40% less than was predicted during its test column evaluation. 3.5. Effect of biolm growth on uidization Microorganisms that metabolize ammonia and other dissolved wastes grow as a biolm attached to the sand surfaces and this attachment prevents the microorganims from being ushed out of the FSB (Cooper and Atkinson, 1981; Shieh et al., 1981; Chang et al., 1991; Summerfelt and Cleasby, 1996; Nam et al., 2000). Cooper and Atkinson (1981), Shieh et al. (1981), and Chang et al. (1991) have modeled biolm growth and substrate uptake in detail. Biolm thickness depends upon a complicated balance between two competing mechanisms: (1) growth rate of the biolm and (2) physical shearing of the biolm (Chang et al., 1991). Growth rate of the biolm depends upon the makeup and age of the microorgan-
ism, as well as the type, concentration, and loading of the growth limiting substrate. The physical shearing of the biolm depends upon the intensity of uid shear, particleparticle collisions, and particlewall collisions (Chang et al., 1991). The process is complicated because of the interdependence of the biolm growth and shearing mechanisms. Biolm growth increases the volume occupied by particles and decreases the effective density of the biolm coated sand, which in turn increases the bed volume and overall expansion of these particles. Increased expansion due to biolm growth can be of special signicance when ne sand (D10 = 0.15 0.3 mm) is used within the FSB, as the biolm thickness may become greater than the diameter of the sand and overall bed expansions of 200% or more can be achieved (Heinen et al., 1996; Tsukuda et al., 1997; Nam et al., 2000; Summerfelt et al., 2004b). In addition, biosolids retention is especially high within FSBs using ne sands, up to 35,000 mg/L of TVS have been measured within the FSB (Tsukuda et al., 1997), because this type of FSB is operating at a relatively low supercial water velocity that does not tend to shear off and washout biolm as would be the case with a larger sand. Portions of biolm that are sloughed from the sand are ushed toward the top of the FSB. With ne sand FSBs, however, the water velocity through the bed is insufcient to lift the larger biolm particles clear of the bed interface and the bed grows deeper with time (Tsukuda et al., 1997; Nam et al., 2000). When necessary, excess biosolids in ne sand FSBs must be siphoned from the top of the bed to prevent the top of the bed from reaching the outlet of the FSB. In contrast, when a larger sand (e.g., D10 > 0.4 mm) is used within a FSB, the supercial water velocity required to maintain bed expansion is possibly 24 times greater than the water velocity required for the ne sand FSB (Tsukuda et al., 1997). Higher water velocities and larger sand grains both increase biolm shear and work to maintain extremely thin biolm coatings on these larger sands, with correspondingly lower TVS concentrations (i.e., 16003000 mg/L) measured in these FSBs (Tsukuda et al., 1997). Biosolids retention in FSBs with larger sands is reduced because the higher water velocity in these vessels tends to carry most of the sloughed biosolids out the top of the biolter (Tsukuda et al., 1997).
291
Fig. 6. Comparison of three sands in expansion tests in a 2.7 m diameter uidized-sand CycloBio1 biolter ( column (^) (from Summerfelt et al., 2004b).
) vs. tests within a 10 cm test
The FSB vertically straties as the largest sand fractions migrate towards the bottom of the bed and the nest sand fractions migrate towards the top of the bed, and because the controlling mechanisms in biolm growth are also vertically stratied (Summerfelt and Cleasby, 1996; Nam et al., 2000). In the lower portion of the FSB, biolm encounters the highest
substrate loading, the largest sand size, the least bed expansion, and the strongest particleparticle/particlewall interactions due to increased turbulence around uid injection sites and due to larger grain sizes and less expansion (Summerfelt and Cleasby, 1996). The upper portion of the FSB has the lowest substrate loading, the smallest grain size, the greatest
292
bed expansion, and the weakest particleparticle/ particlewall interactions. The resulting vertical stratication, at least in ne-sand biolters used in recirculating aquaculture systems, is controlled by physical mechanisms at the bottom of the bed and growth mechanisms in the upper portion of the bed (Summerfelt and Cleasby, 1996). The bottom portion of the FSB consists of relatively larger sand grains that are coated with biolms so thin that they are not visible to the naked eye, whereas, the upper portion of the FSB consists of ner sand with thick biolm and conglomerates consisting of large biosolid particles that may or may not contain sand grains (Summerfelt and Cleasby, 1996; Nam et al., 2000). The middle portion of the FSB can consists of both biolm types, depending upon conditions. In uidized-bed biolters used in aquaculture, a sharp density boundary has often been observed between the two distinct layers, i.e., between the scoured sand layer and the biolmcoated sand layer (Summerfelt and Cleasby, 1996; Nam et al., 2000). At light loads in larval systems, no sludge layer is obtained, even with very ne media (Weaver, 2005).
Once the biolter ow rate (Qbiof, L/min) has been identied and a v0 assumed for a given sand, the crosssectional area (Ab) requirements for the uidized bed can be calculated: Ab Q L=min 1000 cm3 1 min 1 m2 v0 cm=s 1 L 60 s 100 cm2 (12) If necessary, the design can be modied by adjusting tank diameter and sand diameter to provide a tank of convenient size, to allow the use of a graded sand which is available locally, or to increase the potential treatment capacity of the lter (Summerfelt, 1996). The static sand placed into a FSB is generally designed to be 12.5 m deep (38 ft), in part due to practical considerations such as overall vessel height limitations, vessel geometry restrictions, total headloss limits, and bed oxygen demand. Overall expanded bed depth can range from under 2 m to over 5 m, depending upon the situation (Tsukuda et al., 1997; Summerfelt et al., 2004b). However, the sand depth must provide a total bed volume (typically considered as total expanded bed) or a total available surface area sufcient to ensure that 100% of the TAN produced daily can be readily assimilated in the FSB. As summarized by Timmons et al. (2002) and Summerfelt and Vinci (2004), the most important factors in the design of a biolter are (1) the mass of TAN that it removes per day, i.e., the product of the ow rate across the biolter and the change in concentration of ammonia across the biolter; (2) the TAN removal efciency ( f rem) of the biolter. The mass of TAN removed per day can often be increased as the hydraulic loading rate is increased across a biolter. However, increased hydraulic loading rate (i.e., v0 ) across FSBs can decrease the TAN removal efciency as the water retention time is shortened and the mass load of TAN is increased (Tsukuda et al., 1997). The concentration of TAN discharged from a culture tank (TANout, mg/L) is controlled by the f rem of TAN across the biolter, the average daily rate that TAN is produced, i.e., rTAN (kg waste per day), the fraction of water ow that is reused, i.e., R (unitless), and the ow rate of water recirculated through the biolter, i.e., Qbiof (L/min) according to the following
4. Fluidized-sand biolter design and performance criteria FSBs used in recirculating aquaculture systems must be sized to remove 100% of the total ammonianitrogen (TAN) produced daily. A mass balance that accounts for TAN production and removal should be used to determine the rate that water is exchanged through the biolter in order to maintain a TAN concentration within the sh culture tank(s) that is below a specied limit (Timmons et al., 2002; Summerfelt and Vinci, 2004). In most instances, the mass balance indicates that the biolter must treat anywhere from 50 to 100% of the total ow passing through the sh culture tank. The recirculating water that does not have to be treated by the biolter is typically required for transporting additional dissolved oxygen or to provide more dissolved carbon dioxide removal. The size of the biolter (e.g., cross-sectional area and bed depth) and size of the sand selected sets the potential treatment capacity of the biolter. Selection of the sand and desired bed expansion also sets v0 .
293
equation, which was rst developed by Liao and Mayo (1972): 1 TANout 1 R R frem rTAN 106 mg 1 day (13) Qbiof 1 kg 1440 min This equation was derived from mass balances that assume that no waste accumulation can occur in a culture tank, that the make-up water contains no TAN, and that the recirculating system is operating under steady-state conditions, i.e., water ow rates, waste production rates, and unit process treatment efciencies are relatively constant (Timmons et al., 2002). Most recirculating aquaculture systems operating in temperate climates reuse high fractions of their ow (to conserve heated water) and generally operate with only 5100% of the total system water volume exchanged daily, equivalent to a fraction of water ow reused R ! about 0.96 (Timmons et al., 2002; Summerfelt et al., 2004a). In such recirculating systems, waste accumulation depends mainly upon the f rem across the water treatment units and Eq. (14) simples to 1 rTAN 106 mg 1 day TANout frem Qbiof 1 kg 1440 min (14) Eq. (13) or (14) should be used to estimate water ows that are required to achieve the desired TANout, when an accurate estimate of the biolter f rem is available. The biolter treatment efciency that is used during these design calculations should be based upon measurements taken on biolter TAN inlet and outlet concentrations that were collected under conditions that would be similar to those in the design. The following section of this paper will report f rem across FSBs that range from less than 0.1 to greater than 0.9, depending upon the sand size selected, although large number of studies and commercial applications indicate that removal efciency across ne sand FSBs will consistently achieve TAN removal efciencies of >8090%. Other biolter types have their own specic f rem, which will depend upon their design and operating conditions, but f rem typically achieved within other commercial-scale biolter types typically range from 0.1 to 0.5 (Nijhof, 1995; Greiner and
Timmons, 1998; Brazil, 2005). When the TAN concentration in the sh culture tank is held constant, a low f rem results in higher water ow requirements that, even at relatively low head, can produce rather high energy requirements to move sufcient water to meet the TAN concentration limit. Overestimates of the biolter f rem would result in the design of a water recirculating system that could not maintain TAN concentrations within the culture tank within the limits dened in the design. 4.1. Nitrication rate and ammonia removal efciency The total surface area available for microbial attachment is the principle design parameter used to dene the mass of TAN that a biolter can remove daily. However, FSBs present an interesting dilemma regarding the effective use of surface area, as typical lter sands used in FSBs provide specic surface areas of anywhere from 4000 to 20,000 m2/m3. The specic surface area of a loosely-packed bed of static sand (Sb) is relatively large because specic surface area is inversely proportional to the mean diameter of the sand (D50) as described by the equation: Sb 61 e CD50 (15)
where e of a loose packed static bed can be assumed to be 0.45, C can be assumed to be 0.75, and D50 can be estimated using the following equation (J. Cleasby, Iowa State University, Ames, IA, pers. commun.): D50 D10 100:83 log10 UC (16)
In the case of FSBs, it has not been clear whether the nitrication rate on sand is more accurately based on the total expanded bed volume or on the total sand surface area. The specic surface area of a sand bed is so high that the thickness of the biolm coating the sand and the presence of bio-oc particles changes the relationship between TAN treatment capacity and surface area. Therefore, the total volume of the expanded bed will likely provide a better reection of a FSBs treatment capacity than the total surface area within the bed. Sand size controls the beds specic surface area (Eq. (15)), the water velocity required to achieve a
294
Table 4 Average concentrations (S.E.) of dissolved oxygen, total ammonia-nitrogen, and nitrite in the test column inuent and efuents reported by Tsukuda et al. (1997) Parameter Inuent concentration (mg/L) Efuent concentration (mg/L) at effective sand size 0.23 mm Dissolved oxygen Total ammonia-nitrogen Nitrite 10.05 0.16 0.62 0.04 0.038 0.002 5.69 0.19 0.07 0.02 0.067 0.005 0.45 mm 9.08 0.24 0.55 0.04 0.061 0.004 0.60 mm 9.54 0.21 0.57 0.04 0.051 0.004 0.80 mm 9.61 0.15 0.57 0.05 0.045 0.006
given expansion (Eqs. (7), (9) and (10)), the hydraulic retention time through the biolter, the shear forces between biolm-coated sand particles, and the ammonia, oxygen, and organic matter loading rates, which are the product of the biolter inlet concentration times water ow rate. Therefore, in uidized-sand biolters, sand size will also have an effect on the rate and efciency of ammonia, nitrite, and oxygen removal, as well as the biolm thickness and accumulation of biosolids in the bed. The efciency of nitrication per unit surface area is dependent upon the accessibility of the surface to the substrate, the substrate concentration and loading, the mass transfer rate into and out of the biolm, the growth phase of the biolm (lag, log, stationary, and death phases), and by the competition with heterotrophic microbes for space and oxygen (Chang et al., 1991; Manem and Rittman, 1992). Efcient use of the sand surface area is provided in uidized beds by the suspension and rolling of the media grains such that all portions are exposed to the solution. Mass transfer efciency is increased at the biolm surface on the particles within the uidized bed because the high velocities and turbulence required for bed expansion decreases the thickness of the stagnant boundary layer surrounding the biolm. Biolm age can be managed by either introducing clean sand while removing aged biolms from the top of the FSB, or, by selecting a sand diameter that maintains a relatively thin steady-state
biolm, one with no net change in growth or decay. Plug ow of water through the FSB also serves to increase TAN removal efciency in comparison to completely mixed stirred tank reactors, as discussed by Watten and Sibrell (2005). Several studies of FSB nitrication in salmonid recirculating systems and in warm water recirculating system are discussed below. Pilot-scale tests evaluating four sand sizes, Tsukuda et al. (1997). Tsukuda et al. (1997) reports on tests of pilot-scale uidized-sand biolters that were operated in parallel with an established uidized bed sand biolter in a cold-water recirculating system used to produce food-size rainbow trout. An 8-week trial involving twelve 10 cm diameter columns was conducted to determine the nitrication rate and efciency of four different sand sizes, i.e., D10 = 0.23, 0.45, 0.60 and 0.80 mm, that were each replicated three times and that were operated at constant velocities, i.e., 0.82, 1.8, 2.8, and 3.8 cm/s, respectively, for each sand size. Before this study began, the 12 FSBs were given 15 weeks to develop a nitrifying biolm. Change in TAN, nitrite-nitrogen, and dissolved oxygen concentrations were measured across each column during the study (Table 4). The average removal efciency and average removal rate for each of these parameters are reported in Tables 5 and 6. Efuent TAN concentrations leaving the FSBs containing the nest sand were only 0.06 mg/L,
Table 5 Average removal efciencies (S.E.) of dissolved oxygen and total ammonia-nitrogen across the four different sand sizes as reported by Tsukuda et al. (1997) Parameter Removal efciency (%) at effective sand size 0.23 mm Dissolved oxygen Total ammonia-nitrogen 43 1 89 2 0.45 mm 10 1 11 2 0.60 mm 51 82 0.80 mm 41 83
S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302 Table 6 Mean ammonia removal rates (S.E.) across the four different sand sizes as reported by Tsukuda et al. (1997) Total ammonia-nitrogen removal rates At effective sand size 0.23 mm g/day kg/m3 biolm expanded bed per day kg/m3 clean static bed per day g/m2 clean sand per day 8.0 0.5 0.41 0.03 1.5 0.09 0.13 0.01 0.45 mm 2.2 0.3 0.19 0.03 0.48 0.07 0.06 0.01 0.60 mm 2.5 0.5 0.31 0.06 0.45 0.09 0.08 0.02
295
0.80 mm 3.2 1.2 0.39 0.15 0.56 0.21 0.13 0.05
whereas, efuent TAN concentrations leaving the FSBs containing the three largest sands were from 0.55 to 0.57 mg/L (Table 4). This data clearly indicated that the smallest sand size (D10 % 0.23 mm) removed the largest percentage of TAN each pass through the FSB, i.e., 89% removal, whereas the three larger sands only removed 811% of the TAN each pass (Table 5). However, this data did not clearly indicate whether-or-not TAN removal rates were more appropriately expressed in expanded volume or available surface area (Table 6). The TAN removal rate expressed per unit expanded bed volume ranged from 0.19 to 0.41 kg ammonia/m3/day across all four sand sizes (Table 6). Full-scale study on nitrication in a FSB operated with or without an internal biolm stripping mechanism (unpublished). Biolm stripping devices have been used to shear thick biolm from sand since the early 1970s in commercial wastewater treatment units (Sutton and Mishra, 1991). However, control of biolter bed expansion in ne-sand FSBs found in recirculating aquaculture systems is typically achieved by simply siphoning off excess bed growth when it comes within 0.30.6 m of the vessel outlet. In a 310 day unpublished research study conducted at TCFFI, S. Summerfelt, M. Durant, and D. Bullock evaluated a mechanism for stripping thickening biolms within a FSB to control bed growth and reduce the size of biosolids contained within the bed. A horizontal pipe manifold type FSB was tested that was 1.52 m diameter by 2.44 m tall. The FSB was operated in a recirculating salmonid production system that was previously described by Heinen et al. (1996). During this study, the estimated trout biomass in the recirculating system averaged 814 kg (ranging from 684 to 981 kg due to stocking and harvesting events) and the mass of feed fed weekly averaged 174 kg. A pump was used to force the
biosolid particles that collect near the top of the bed to near the bottom of the bed where the uid shear was greatest. A small magnetic drive submersible pump (Little Giant) was used to move the relatively larger bio-particles from near the top of the bed to its base, which was in the scoured-sand region of the bed. The magnetic drive submersible pump did not wear-out from sand abrasion during 230 days of use. The ammonia, nitrite, and oxygen concentrations in the biolter inlet and outlet ow were measured approximately three times weekly. Sand samples were collected weekly from the biolter at three different depths and these samples were analyzed for size of clean sand and overall diameter of the larger bio-particles (Table 7) which look similar to large brown cottage cheese curds contained in the samples using a digital image analysis software package (Mocha1, Jandel Scientic, San Rafael, CA). Data collected during the period before the biolm stripping mechanism was added to the biolter was compared to data collected during operation of the biolm stripping mechanism (Table 7). Results from this study indicate that the FSB was vertically stratied with respect to sand size, bio-particle size, and expansion (Table 7). Results also indicate that the diameter of the larger bio-particles could be reduced by operating the biolm stripping mechanism (Table 7). Pumping the biosolids from the top of the biolter bed to the bottom of the bed must have increased the opportunity for the biolm to shear because of the more intense physical interactions produced by the pump impeller, vessel bottom, and the larger sand grains located in the lower portion of the bottom. The turbulent conditions and water jets in the water injection regions at the bottom of the sand bed were also thought to contribute favorably to biolm scouring. The combined depth of the sand and biosolids in the biolter averaged 1.93 0.01 m
296
Table 7 Diameter (mean S.E.) of cleaned sand and of bio-oc particles collected from the upper, middle, and lower portions of the FSB Mean sand size (mm) Stripper off Upper bed Middle bed Lower bed Stripper on Upper bed Middle bed Lower bed 0.289 0.004 0.319 0.005 0.420 0.004 0.305 0.004 0.330 0.003 0.430 0.004 Bio-oc diameter (mm) 1.096 0.077 1.706 0.114 None present 0.887 0.044 0.936 0.031 None present Bed expansion (%) 257 18 203 13 59 7 231 20 207 11 68 4
In addition, approximate bed expansions are reported for samples taken from upper, middle, and lower regions of the biolter and uidized in 10.2 cm diameter test columns at a supercial velocity of 0.63 cm/s.
(76 1 in.) and 1.76 0.03 m (69 1 in.), respectively, during the control period and the period when the biolm stripping mechanism was activated. However, the biolm stripping mechanism was able to maintain a fairly constant bed depth without the having to siphon biosolids from the top of the bed, which was required when the biolm stripping mechanism was not in use. The TAN and nitrite-nitrogen concentrations exiting the FSB biolter were both low, i.e., TAN was 0.08 and 0.12 mg/L and nitrite-nitrogen was 0.04 and 0.07 mg/L, when operated with or without the biolm stripping mechanism, respectively (Table 8). In addition, TAN removal efciency and removal rate were comparable, i.e., removal efciency of 88 and 82% and removal rate of 164 and 146 g TAN removed per day per cubic meter of expanded bed, when operated with or without the biolm stripping mechanism, respectively (Table 8).
After completing this study, the authors concluded that, although the biolm stripping mechanism was effective at controlling bed expansion and did not reduce the performance of the FSB, it was considered undesirable to increase biolm shearing and washout of ne biosolids from the FSB because this in turn would increase the number of ne solids coming in contact with sh gills. Producing ne solids within a water recirculating system may not be of much concern when the sh species cultured can withstand high levels of suspended solids. Full-scale study on nitrication in a CycloBio1 FSB, Summerfelt et al. (2004b). Nitrication within a 2.74 m (9 ft) diameter 6.0 m (20 ft) tall CycloBio1 FSB in a recirculating salmonid system at TCFFI (described by Summerfelt et al., 2004a) was evaluated during 2 years of operation using a Parry Company 35/ 42 Richmond Dale silica sand, i.e., D10 0.23 mm, and over 8 months of operation using the US Silica
Table 8 Total ammonia-nitrogen (TAN), nitrite-nitrogen, and oxygen biolter inlet and outlet concentrations, as well as biolter removal efciencies and removal rates within a FSB operated with or without a biolm shearing mechanism Biolter inlet (mg/L) Stripper off TAN Nitrite-nitrogen Oxygen Stripper on TAN Nitrite-nitrogen Oxygen 0.63 0.02 0.10 0.02 10.0 0.2 0.66 0.01 0.06 0.00 9.4 0.3 Biolter outlet (mg/L) 0.12 0.02 0.07 0.02 6.2 0.2 0.08 0.01 0.04 0.00 4.5 0.2 Removal efciency (%) 82 2 90 2 38 2 88 1 94 0 50 3 Removal rate (g/day/m3 of expanded bed) 146 8 157 9 1070 80 164 5 169 5 1380 140
S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302 Table 9 Nitrication performance using two different sands in a full-scale CycloBio1 FSB as reported by Summerfelt et al. (2004b) #1 Mapleton sand 1.1% makeup Total ammonia-nitrogen Biolter inlet (mg/L) Biolter outlet (mg/L) Biolter delta TAN (mg/L) TAN removal efciency (%) TAN removal rate, g/day/m3 of expanded bed depth Nitrite-nitrogen Biolter inlet (mg/L) Biolter outlet (mg/L) Dissolved carbon dioxide Biolter inlet (mg/L) Biolter outlet (mg/L) Biolter delta CO2 (mg/L) Dissolved oxygen Biolter inlet (mg/L) Biolter outlet (mg/L) Biolter delta O2 (mg/L) Dissolved organic carbon Biolter outlet (mg/L) Makeup ow, % total ow Feed (kg/day) Flow through biolter (L/min) Biolter bed depth (m) 1.49 0.04 0.22 0.01 1.24 0.03 83.1 0.2 160 0.28 0.03 0.25 0.07 25 0 28.5 2 42 9.4 0.1 2.5 0.0 6.9 0.2 7.4 0.1 1.1 0.0 126.4 3.1 2696 17 5.27 0.01 6.4% makeup 1.18 0.04 0.09 0.01 1.09 0.03 92.2 0.7 140 0.06 0.00 0.02 0.00 20 2 28 2 81 10.9 0.4 4.6 0.1 6.4 0.3 2.7 6.4 0.1 136.7 11.7 2716 6 5.27 0.01 35/42 Richmond Dale sand 3.4% makeup 1.68 0.02 1.02 0.05 0.660 0.066 39.3 3.5 170 0.72 0.02 0.74 0.01 8.1 0.1 3.4 0.1 143.2 4.6 4497 11.5 4.37 0.01
297
8.4% makeup 1.07 0.11 0.38 0.01 0.69 0.10 64.2 2.9 170 0.27 0.02 0.31 0.01 4.1 0.7 8.4 0.3 152.7 0 4447 0 4.37 0.01
Companys #1 Mapleton silica sand, i.e., D10 0.18 mm (Summerfelt et al., 2004b). In the rst nitrication study, the CycloBio1 FSB contained the coarser of the two sands (e.g., the 35/42 Richmond Dale sand) and the entire recirculating water ow, approximately 45004800 L/min, was pumped to the base and up through the top of CycloBio1 uidizedsand biolter. In the second nitrication study, the CycloBio1 contained the ner of the two sands (e.g., the #1 Mapleton sand) and only 60% of recirculating water ow, approximately 2700 L/min, was pumped through the vessel (Table 9). The remaining 40% of the ow by-passed the CycloBio1 and owed directly to the top of the cascade aeration column. The initial overall clean sand bed expansion was approximately 40% in the rst study and approximately 60% in the second study. However, after a biolm was established, overall bed depth was allowed to grow to 4.37 m (190% expansion) and 5.27 m (216% bed expansion), respectively, for the coarser (35/42 Richmond Dale) sand and the ner (#1 Mapleton)
sand. Bed depth was maintained at or below these levels by continuously siphoning biolm from the top of the bed. The v0 through the CycloBio1 was approximately 1.36 cm/s (20 gpm/ft2) and 0.77 cm/s (11 gpm/ft2) for the coarser (35/42 Richmond Dale) sand and the ner (#1 Mapleton) sand, respectively. With the ner (#1 Mapleton, D10 = 0.18 mm) sand, the uidized-sand CycloBio1 biolter maintained >8090% TAN removal efciency each pass through the biolter (Table 9). Typical makeup water ow rates were 48% of total recirculating ow and under these conditions TAN removal efciencies averaged 92.2 0.7% and nitrite-nitrogen concentrations averaged 0.06 0.00 mg/L (Table 9). When makeup water ows were reduced to approximately 1% of the total recirculating ow, dissolved organic carbon concentrations increased to approximately 7.4 0.05 mg/L, the TAN removal efciency dropped to 83.1 0.2%, and nitrite-nitrogen concentrations rose to 0.28 0.03 mg/ L (Table 9). Dissolved oxygen was never limiting within the biolter, as total dissolved oxygen con-
298
sumption across the biolter containing the ner sand averaged 6.46.9 mg/L, when it was supporting 126 137 kg/day feed loading on the system (Table 9). With the coarser (35/42 Richmond Dale, D10 0.23 mm) sand, the uidized-sand CycloBio1 biolter typically maintained TAN removal efciencies of 64.2 2.9% and nitrite-nitrogen concentrations of 0.31 0.01 mg/L when supporting a relatively high feed loading rate (152.8 kg/day). When makeup water ows were reduced to approximately 3.4% of the total recirculating ow, dissolved organic carbon concentrations increased to approximately 8.1 0.1 mg/L, the TAN removal efciency dropped to 39.3 3.5%, and nitrite-nitrogen concentrations rose to 0.74 0.01 mg/L (Table 9). This data indicates that, as the recirculating system makeup ow was restricted, the resulting accumulation in dissolved organic carbon concentration decreased TAN removal efciencies and increased the steady state concentrations of TAN and nitritenitrogen, especially when the FSB was not operated with the more efcient of the two sand sizes that were evaluated. However, TAN removal rate ranged from 140 to 170 g/day/m3 of expanded bed depth over all conditions tested (Table 9). Summary of nitrication results. Data from the two full-scale FSB studies described above (Tables 8 and 9) indicate that TAN removal efciencies depend upon sand size (Tsukuda et al., 1997; Summerfelt et al., 2004b). These studies indicated that TAN removal efciencies of >8090% could be achieved when a ne sand was used (i.e., v0 % 0:630:77 cm=s), but that TAN removal efciency declined sharply with larger sand sizes. Data from the two full-scale FSB studies described above (Tables 8 and 9) also indicate that TAN removal rates of 140170 g/day/m3 of expanded bed depth have been consistently provided under widely different conditions in full-scale FSB contained in coldwater recirculating systems. TAN removal rates as high as 400 g/day/m3 of expanded bed depth were achieved in test columns. Thomasson (1991), Monaghan et al. (1996), and Shea et al. (1997) have studied nitrication on larger sands in FSBs found in warmwater systems. In warm water systems, TAN removal rates range from 0.6 to 1.0 kg/day/m3 expanded bed volume (Timmons and Summerfelt, 1998). The optimum sand size for use in warm-water systems (i.e., at 2530 8C) appears to be closer to 0.50.7 mm,
e.g., ranging from a 20 to 40 mesh sand to a 1630 mesh sand, where these sands are expanded approximately 50% with water velocities v0 of 2.0 and 3.2 cm/s, respectively (Timmons and Summerfelt, 1998). 4.2. Carbon dioxide production and dissolved oxygen consumption Nitrifying bacteria and heterotrophic microorganisms in a FSB will respire and produce a net increase in dissolved carbon dioxide and a net decrease in dissolved oxygen. The production of dissolved carbon dioxide within the biolter can be estimated from measurements of the concentrations of dissolved oxygen and TAN removed from the water owing through the biolter. Approximately 5.9 mg/L of CO2 is produced and approximately 4.6 mg/L of dissolved oxygen is consumed for every 1 mg/L of TAN consumed across a submerged biolter (Summerfelt and Sharrer, 2004). In addition, approximately 1.38 mg/L of CO2 are produced for every 1 mg/L of dissolved oxygen consumed (Summerfelt and Sharrer, 2004). During the Summerfelt and Sharrer (2004) study, the FSB at TCFFI produced 4.1 0.2 mg/L of carbon dioxide while removing 0.51 0.02 mg/L TAN and removing 3.8 0.2 mg/L dissolved oxygen. During the same period the sh produced 6.9 0.4 mg/L of carbon dioxide within this same recirculating ow, so the FSB accounted for approximately 37% of the total carbon dioxide concentration produced within this recirculating salmonid system. Across a more heavily loaded FSB described by Summerfelt et al. (2004b), the dissolved carbon dioxide production has averaged as much as 8 mg/L (Table 9). In addition, dissolved oxygen concentrations exiting the FSBs are relatively low, ranging from 2.5 to 6.2 mg/L (Tables 4, 8 and 9) within the studies summarized above. Approximately 60% of the dissolved oxygen consumed in the FSB operated in a coldwater recirculating system goes towards nitrication (Summerfelt and Sharrer, 2004). Therefore, for every 1 mg/L of TAN removed across the FSB, approximately 7.7 mg/L (=4.6 mg/L DO/0.6) of dissolved oxygen will be consumed, which assumes that 60% of the dissolved oxygen consumed went towards TAN removal. Consequently, oxygen limitations within FSBs can become a problem when the desired removal of TAN is high, i.e., greater than 0.8
299
1.2 mg/L across the FSB, depending upon concentration of dissolved oxygen entering the FSB that was assumed to be the saturation concentration in cold and warm water applications, respectively. According to more traditional wastewater treatment biolter technologies discussed by Zhu and Chen (2002), dissolved oxygen will begin to limit TAN removal in a submerged biolter when the ratio of dissolved oxygen to TAN concentration becomes less than 1.52.0. Therefore, if the TAN concentration exiting the FSB was 1.0 mg/L, then at least 1.52.0 mg/L of dissolved oxygen would have to be present to prevent oxygen from limiting nitrication. In the ne sand FSB applications described above, TAN outlet concentrations were less than 0.3 mg/L, so dissolved oxygen concentration would not be limiting until dissolved oxygen concentrations approached 0.6 mg/L. To counter the dissolved carbon dioxide production and dissolved oxygen consumption across the FSB and also across the sh culture tank(s) within this recirculating system a forced-ventilated cascade aeration column is typically placed immediately after the FSB to reduce dissolved carbon dioxide concentrations and increase dissolved oxygen concentrations to near saturation (Summerfelt et al., 2003; Summerfelt and Sharrer, 2004).
5. Fluidized-sand biolter operation and management practices 5.1. Managing biolm growth Biological growth occurs within FSBs, which results in a growing bed that will probably not reach equilibrium before some form of biolm management must be instituted to avoid washout of lightened biolm-coated sand (Tsukuda et al., 1997). A disengagement zone between the interface at the top of the expanded bed and the FSB outlet of at least 0.31.0 m is necessary to reduce biosolids washout through the FSB outlet. Bio-bed growth is especially common in ne-sand FSBs (i.e., with a D10 less than about 0.3 mm), because these biolters operate at a v0 that is insufcient to lift the larger biolm particles clear of the vessel and the bed grows deeper with time (Bullock et al., 1993; Heinen et al., 1996; Summerfelt and Cleasby, 1996; Tsukuda et al., 1997;
Timmons and Summerfelt, 1998; Nam et al., 2000; Summerfelt et al., 2004b). When D10 exceeds 0.6 mm, biosolids do not accumulate within the expanded sand bed, but detached biolm particles can sometimes collect in a distinct layer above the expanded sand layer (Tsukuda et al., 1997; Timmons and Summerfelt, 1998). The expanded bed of biolm coated sand is uid, therefore a siphon withdrawing ow and biosolids from one point in the biolter can remove all uidized biosolids at depths above this level (for all sand sizes). Thus, bed expansion can be managed by removing biolm-coated sand or by stripping the biolm off the sand by increasing the physical shear forces. Siphoning the biosolids from the top of the expanded bed is a relatively simple and frequently used technique to manage bed expansion (Bullock et al., 1993; Heinen et al., 1996; Summerfelt and Cleasby, 1996; Tsukuda et al., 1997; Timmons and Summerfelt, 1998; Summerfelt et al., 2004b). Siphoning the biosolids layer that collects above an expanded bed of relatively large sand (i.e., with a D10 of 0.6 mm or larger) is especially simple, because the expansion depth of these sands remains fairly constant and the biosolids can be removed relatively free from sand (Tsukuda et al., 1997; Timmons and Summerfelt, 1998). However, when biosolids are siphoned from beds containing a ner sand, then some sand will be lost when the biosolids are removed. With ner sands, the sand lost during biosolids removal must be replaced within typically 12 years. Yet, the cost of sand replacement is relatively low. For example, roughly 50% of the ne-sand in a large-scale CycloBio1 FSBs at TCFFI was missing after approximately 2 years of operation. Replacement of this sand would cost $ 500 750 in this biolter, which was sized to treat the TAN produced by feeding 200 kg daily. However, to prevent the nes contained in the new sand from contacting sh, ideally new sand would only be installed in a recirculating system that does not contain sh or the new sand should be pre-ushed to remove the nes. Researchers at TCFFI have also investigated controlled biolm thickness by shearing the biolm in-vessel using a pump to transport the occulant particles from the top of the biolter to the bottom of the bed, where shear forces are greatest. The biolm stripping system effectively maintained bed
300
expansion at a xed level and reduced biolm thickness without trading-out sand (unpublished data). However, it was concluded that increasing biolm shearing and washout of ne biosolids from the FSB was undesirable, because this in turn would increase the number of ne solids coming in contact with sh gills. 5.2. Managing ow and avoiding bubbles A FSB must be operated within a fairly narrow water ow range, i.e., within about 1030% of its design ow, in order to maintain proper bed expansion. In addition, water ow through a FSB cannot cease for more than approximately 624 h, depending upon conditions, in order to prevent anaerobic condition from occurring that could cause a signicant loss in the nitrication capacity of the FSB. When ow is resumed after the FSB has been down for more than 12 h, then the water that is initially ushed can contain elevated concentrations of TAN and suspended solids. Therefore, it is a good management practice to directly discharge the ow of water that was initially ushed through the FSB or let the water recycle while by-passing all of the sh culture tanks assuming that the recirculating system design allows for this option. Bubbles must also be prevented from entering the FSB, which requires preventing bubbles from entering through leaky pipe ttings or through vortexing or bubble entrainment in the pump sump. Bubbles can create serious problems in FSBs, as they oat out sand and biosolids. 5.3. Determining effectiveness of bed expansion Upon rst expansion of the FSB, and then again as often as once a month or as infrequently as biannually, the effectiveness of sand bed expansion should be determined by probing the base of the lter with a pole to establish the depths and locations of sand piles, if any are present. Determining whether the sand bed is uniformly expanded, or not, is necessary to trouble shoot the FSB and determine if the distribution manifold is becoming plugged. If sand mounds taller than about 0.3 m are detected, then it may be necessary to ush debris from the ow distribution manifold.
Depth of the scoured-sand layer should also be determined as an indicator of the total sand available within the bed. 5.4. Mechanisms to unplug the ow distribution manifold Cleanout of the ow distribution manifold is rarely required. However, cleanout of the ow distribution manifold is necessary in the rare event when the manifold becomes plugged with sand or other debris (Summerfelt et al., 1996), which is most likely to occur if the backow prevention device is by-passed or malfunctions. At least three of the ve types of the FSBs described above (Fig. 1) provide cleanout mechanisms to unplug the ow distribution manifold, i.e., the vertical probe FSB, the horizontal pipe manifold FSB, and the CycloBio1 FSB. These three cleanout mechanisms have all been evaluated at TCFFI over the past decade. As an example, a blind ange covering an access port into the annular space of a CycloBio1 FSB can be removed3 to allow a short burst of pumped water to ush sand from that region of the annular space. Then, after the access port has been resealed with the blind ange, the CycloBio1 FSB has been found to re-uidize and clear the remaining sand from the annular chamber (TCFFI, unpublished data). As another example, the lateral pipes in a horizontal pipe manifold can be readily ushed of any plugging debris (Summerfelt et al., 1996) by ushing water down the pipe after the screw caps are removed from the overhead tees at the end of each plugged lateral (Fig. 1). More pressure can be applied to individual laterals by isolating them with valves and applying the entire ow to the obstructions. In addition, a hose supplying pressurized water can be inserted through a clean-out port at the top of each injection pipe (Fig. 1) to backwash any sand or debris from the vertical pipe. Acknowledgements This work was supported by the United States Department of Agriculture, Agricultural Research Service, most recently under grant agreement number
Note that the blind ange over the access port into the annular chamber must not be removed until after water inside the vessel has been removed.
3
301
59-1930-1-130. I thank Susan Glenn, Thomas Waldrop, Mark Sharrer, John Davidson, Christine Marshall, Daniel Bullock, Graham Bullock, Scott Tsukuda, and Martin Durant for their assistance with the biolter research. The experimental protocol and methods used in TCFFI biolter studies were in compliance with Animal Welfare Act (9CFR) requirements and are approved by the Freshwater Institute Institutional Animal Care and Use Committee. References
American Society for Testing and Materials (ASTM), 1985. Annual Book of ASTM Standards, vol. 04.02, Concrete and Mineral Aggregates ASTM, Philadelphia, PA. American Water Works Association (AWWA), 1971. Water Quality and Treatment, 3rd ed. McGraw-Hill, New York, NY. American Water Works Association (AWWA), 1990. Water Treatment Plant Design, 2nd ed. McGraw-Hill, New York, NY. American Water Works Association (AWWA), 2004. Filter Media Sand. AWWA Sourcebook AWWA, Denver, CO, p. 151. Brazil, B.L., 2005. Performance and operation of a rotating biological contactor in a tilapia recirculating aquaculture system. Aquacult. Eng. 34 (3), 261274. Bullock, G., Hankins, J., Heinen, J., Starlipper, C., Teska, J., 1993. Quantitative and qualitative bacteriological studies on a uidized sand biolter used in a semi-closed trout culture system. Biological Report 17 U.S. Fish and Wildlife Service, Washington, D.C., 15 pp. Chang, H.T., Rittman, B.E., Amar, D., Heim, R., Ehlinger, O., Lesty, Y., 1991. Biolm detachment mechanisms in a liquid-uidized bed. Biotech. Bioeng. 38, 499506. Cleasby, J.L., 1990. Filtration. In: Pontius, F.W. (Ed.), Water Quality and Treatment. 4th ed. American Water Works Association/ McGraw-Hill, New York, pp. 455560. Cooper, P.F., Atkinson, B., 1981. Biological Fluidized Bed Treatment of Water and Wastewater. Water Research Centre EllisHorwood, Chichester, UK. Denn, M.M., 1980. Process Fluid Mechanics. Prentice-Hall, Englewood Cliffs, NJ. Dharmarajah, A.H., Cleasby, J.L., 1986. Predicting the expansion of lter media. J. Am. Water Works Assoc. 78 (12), 6676. Ergun, S., 1952. Fluid ow through packed columns. Chem. Eng. Prog. 48 (2), 89. Fan, K.-S., 1981. Sphericity and uidization of granular lter media. Masters Thesis, Iowa State University, Ames, IA. Forsythe, A., Hosler, K.C., 2002. Experiences in constructing and operating cold water recirculating aquaculture facilities for salmon smolt production. In: Rakestraw, T.T., Douglas, L.S., Flick, G.J. (Eds.), Proceedings of the Fourth International Conference on Recirculating Aquaculture, Virginia Polytechnic Institute and State University, Roanoke, VA, pp. 325334. Goldman, J.N., Rosenau, J.R., 2000. Reissue of US Patent 5,330,652 (RE36,660). Fluidized bed reactor and distribution system.
United States Patent and Trademark Ofce, Alexandria, Virginia. Geldart, D., 1990. Estimation of the basic particle properties for use in uid-particle process calculations. Powder Technol. 60, 113. Greiner, A.D., Timmons, M.B., 1998. Evaluation of the nitrication rates of microbead and trickling lters in an intensive recirculating tilapia production facility. Aquacult. Eng. 18, 189200. Heinen, J.M., Hankins, J.A., Weber, A.L., Watten, B.J., 1996. A semi-closed recirculating water system for high density culture of rainbow trout. Prog. Fish Culturist 58, 1122. Holder, J., 2002. Retrots of ow through to reuse/recirculation technology. In: Rakestraw, T.T., Douglas, L.S., Flick, G.J. (Eds.), Proceedings of the Fourth International Conference on Recirculating Aquaculture, Virginia Polytechnic Institute and State University, Roanoke, VA, pp. 335345. Jewell, W.J., 1990. Fundamentals and advances in expanded bed reactors for wastewater treatment. In: Tyagi, R.D., Vembu, K. (Eds.), Wastewater Treatment by Immobilized Cells. CRC Press, Boca Raton, FL, pp. 223252. Liao, P.B., Mayo, R.D., 1972. Salmonid hatchery water reuse systems. Aquaculture 1, 317335. Malone, R.F., Burden, D.G., 1988. Design of Recirculating Soft Crawsh Shedding Systems. Louisiana Sea Grant College Program, Center for Wetland Resources, Louisiana State University, Baton Rouge, LA. Manem, J.A., Rittman, B.E., 1992. The effects of uctuations in biodegradable organic matter on nitrication lters. J. Am. Water Works Assoc. 84 (4), 147151. Monaghan, T.J., Delos Reyes, A.A., Jeansonne, T.M., Malone, R.F., 1996. Effects of media size on nitrication in uidized sand lters. In: Aquaculture America96 Book of Abstracts, World Aquaculture Society, Baton Rouge, LA, p. 110. Montagne, M., 2004. The evolution of two warm water recirculating hatcheries used for propogation of endangered species in the upper Colorado River Drainage System. In: Rakestraw, T.T., Douglas, L.S., Correa, A., Flick, G.J. (Eds.), Proceedings of the Fifth International Conference on Recirculating Aquaculture, Virginia Polytechnic Institute and State University, Roanoke, VA, pp. 273279. Montgomery, J.M., Consulting Engineers, Inc., 1985. Water Treatment Principles and Design. John Wiley and Sons, New York. Nam, T.K., Timmons, M.B., Montemagno, C.D., Tsukuda, S.M., 2000. Biolm characteristics as affected by sand size and location in uidized bed vessels. Aquacult. Eng. 22, 213224. Nijhof, M., 1995. Bacterial stratication and hydraulic loading effects in a plug-ow model for nitrifying trickling lters applied in recirculating sh culture systems. Aquaculture 134, 49 64. Shea, G.O., Timmons, M.B., Summerfelt, S.T., Tsukuda, S., 1997. Characterization of uidized sand beds for warm water systems. In: World Aquaculture97 Book of Abstracts, World Aquaculture Society, Baton Rouge, LA, p. 421. Shieh, W.K., Sutton, P.M., Kos, P., 1981. Predicting reactor biomass concentration in a uidized-bed system. J. Water Pollut. Control Federation 53, 15741584. Summerfelt, S.T., 1996. Engineering design of a water reuse system. In: Summerfelt, R.C. (Ed.), Walleye Culture Manual, NRAC
302
S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302 Aquaculture, Virginia Polytechnic Institute and State University, Roanoke, VA, pp. 342354. Timmons, M.B., Helwig, N., Summerfelt, S.T., 2000. The Cyclone sand biolter: a new design concept and eld evaluation. In: Libey, G.S., Timmons, M.B. (Eds.), Proceedings of the Third International Conference on Recirculating Aquaculture. Virginia Polytechnic Institute and State University, Roanoke, VA, pp. 222226. Timmons, M.B., Ebeling, J.M., Wheaton, F.W., Summerfelt, S.T., Vinci, B.J., 2002. Recirculating Aquaculture Systems, 2nd ed. Cayuga Aquaculture Ventures, LLC, Ithaca, NY. Tsukuda, S.M., Hankins, J.A., Marshall, C.P., Summerfelt, S.T., Bullock, G.L., Sawyer, T.K., 1997. Effects of sand size on uidized-bed biolter performance in cold water systems. In: Timmons, M.B., Losordo, T. (Eds.), Advances in Aquacultural Engineering, NRAES-105. Northeast Regional Agricultural Engineering Service, Ithaca, NY, pp. 368 380. Watten, B.J., Sibrell, P.L., 2005. Comparative performance of xedlm biological lters: application of reactor theory. Aquacult. Eng. 34 (3), 198213. Weaver, D.E., 1991. Performance of ne sand uidized bed. In: Proceedings of the Design of High-density Recirculating Aquaculture Systems. Louisiana Sea Grant College Program Publication, Louisiana State University, Baton Rouge, LA, pp. 18 23. Weaver, D.E., 2005. Design and operations of ne media uidized bed biolters for meeting oligotrophic water requirements. Aquacult. Eng. 34 (3), 303310. Weber, W.J., 1972. Physicochemical Processes for Water Quality Control. John Wiley and Sons, New York. Wen, C.Y., Yu, Y.H., 1966. Mechanics of uidization. In: Chemical Engineering Progress Symposium Series 62, American Institute of Chemical Engineers, New York. Wilton, S.J., 2002. Design of a 50 T Salmon Smolt Operation. In: Rakestraw, T.T., Douglas, L.S., Flick, G.J. (Eds.), Proceeding of the Fourth International Conference on Recirculating Aquaculture, Virginia Polytechnic Institute and State University, Roanoke, VA, pp. 325334. Zhu, S., Chen, S., 2002. The impact of temperature on nitrication rate in xed lm biolters. Aquacult. Eng. 26, 221237.
Culture Series 101. Central Regional Aquaculture Center Publication Center, Iowa State University, Ames, IA, pp. 277309. Summerfelt, S.T., Cleasby, J.L., 1996. A review of hydraulics in uidized-bed biological lters. Trans. Am. Soc. Agric. Eng. 39 (3), 11611173. Summerfelt, S.T., Sharrer, M.J., 2004. Design implication of biolter carbon dioxide production within recirculating salmonid culture systems. Aquacult. Eng. 32, 171182. Summerfelt, S.T., Vinci, B.J., 2004. Avoiding water quality failures. Part 2. Recirculating systems. World Aquacult. 35 (4) 911, 71. Summerfelt, S.T., Wade, E.M., 1998. Fluidized-sand biolters installed at two farms. Recirc Today 1 (1), 1821. Summerfelt, S.T., Hankins, J.A., Durant, M.D., Goldman, J.N., 1996. Removing obstructions: modied pipe-lateral ow distribution mechanism reduces backow in uidized-sand biolters. Water Environ. Technol. 8 (11), 3949. Summerfelt, S.T., Bebak-Williams, J., Tsukuda, S., 2001. Controlled systems: water reuse and recirculation. In: Wedemeyer, G. (Ed.), Fish Hatchery Management. 2nd ed. American Fisheries Society, Bethesda, MD, pp. 285395. Summerfelt, S.T., Davidson, J.T., Waldrop, T., Vinci, B.J., 2003. Evaluation of full-scale carbon dioxide stripping columns in a coldwater recirculating system. Aquacult. Eng. 28, 155169. Summerfelt, S.T., Wilton, G., Roberts, D., Savage, T., Fonkalsrud, K., 2004a. Developments in recirculating systems for arctic char culture in North America. Aquacult. Eng. 30, 3171. Summerfelt, S.T., Davidson, J., Helwig, N., 2004b. Evaluation of a full-scale CycloBio uidized-sand biolter in a coldwater recirculating system. In: Rakestraw, T.T., Douglas, L.S., Correa, A., Flick, G.J. (Eds.), Proceedings of the Fifth International Conference on Recirculating Aquaculture, Virginia Polytechnic Institute and State University, Roanoke, VA, pp. 227237. Sutton, P.M., Mishra, P.N., 1991. Biological uidized beds for wastewater treatment: a state-of-the-art review. Water Environ. Technol. 3 (8), 5256. Thomasson, M.P., 1991. Nitrication in uidized-bed sand lters for use in recirculating aquaculture systems. Masters Thesis. Louisiana State University, Baton Rouge, LA. Timmons, M.B., Summerfelt, S.T., 1998. Application of uidizedsand biolters. In: Libey, G.S., Timmons, M.B. (Eds.), Proceedings of the Second International Conference on Recirculating

Design and Management of Fluidized-Sand Bio Filters

Caricato da

Informazioni sul documento

Descrizione originale:

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Design and Management of Fluidized-Sand Bio Filters

Caricato da

Copyright:

Formati disponibili

Aquacultural Engineering 34 (2006) 275302 www.elsevier.

Design and management of conventional uidized-sand biolters

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

0:56543 1:09348 logRe1 0:17971logRe1 2 0:00392logRe1 4 1:5logc2

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

Bulk density of the bed, rb

Particle size distribution

Effective size, D10

Calculating size, D90

Equivalent diameter, Deq

Particle specic surface area, Sp

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

) vs. tests within a 10 cm test

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

0.80 mm 3.2 1.2 0.39 0.15 0.56 0.21 0.13 0.05

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

S.T. Summerfelt / Aquacultural Engineering 34 (2006) 275302

Potrebbero piacerti anche